Conditional on having surgery, there appear to be some advantages associated with being operated on in hospitals that are categorized as having a more aggressive treatment style. Our results show that patients in more aggressive hospitals did not develop an increase in complications, and if a complication did occur, patients were significantly more likely to survive these events than in less aggressive hospitals.
Our results do not suggest that aggressiveness is unambiguously good, as we have focused on only one of three possible consequences of a more aggressive style—that being surgical quality of care (see Equations (1)
). Aggressiveness may have adverse effects that operate through the other two channels: the probability of death for nonsurgical patients and the incidence of surgery. However, to make the case for reducing aggressiveness, any adverse effects that work through these other two channels must offset some of the positive effects suggested by our results.
Because aggressiveness does not influence the rate of complications, but does improve failure-to-rescue, we may ask if better survival in aggressive hospitals is a good thing. While we cannot assume that lower failure-to-rescue in aggressive hospitals is always desirable, it is a much more reasonable assumption than to believe that the increase in saved lives associated with higher aggressiveness is unfortunate—that a patient who develops a complication and is saved would have been better off dead. We found evidence from that the better 30-day mortality associated with patients treated at more aggressive hospitals did not produce elevated death rates after 30 days in those survivors treated at the more aggressive hospitals as compared with 30-day survivors in less aggressive hospitals. Hence, the benefits associated with more aggressive hospitals appear durable.
Our study has several limitations. Our data were limited to Medicare claims and we did not have the ability to collect chart information. There may be selection of less severely ill patients in the hospitals with more aggressive treatment style that may account for our findings. We believe there is no indication that inadequate adjustment somehow led to the observed association between increased aggressiveness and better outcomes. All analyses, even ones with only adjustment for procedure, provided similar results. Second, when we calculated the probability of death following complications (the failure-to-rescue analyses), we again found similar results. As the failure-to-rescue analyses only include those with complications, the severity-adjustment problem is reduced because complicated patients are more homogeneously ill than a mixture of patients with and without complications (Silber, Rosenbaum, and Ross 1995a
; Silber et al. 2007
;). Finally, the stability analyses on a subset of patients having low-variation procedures also displayed similar findings. Nevertheless, we cannot rule out the possibility that hospitals deemed to be more aggressive through the aggressiveness measure are the very same hospitals that over-report complications and comorbidities. If this were the case, a spurious association may be observed between increased aggressiveness and lower adjusted mortality.
Another limitation involves our ability to be certain that aggressive treatment style measures as reflected in the Dartmouth Atlas for the nine conditions used to construct their measure are reasonable proxies for surgical care aggressiveness in the procedures we report on in this study. As evidence that the Dartmouth aggressiveness measure does reflect aggressive hospital style beyond the nine conditions used to define the measure, we have found that these aggressiveness measures are associated with total spending on surgical patients (Kaestner and Silber, unpublished data, 2009). Specifically, aggressiveness measures for a hospital were strongly and positively associated with resource use for patients admitted to that hospital for general surgery, orthopedic surgery, and vascular surgery. In other words, the same style that leads to high expenditure in the Dartmouth Atlas also leads to high expenditure on our surgical admissions within the same hospital. In Table 5 of Appendix SA2
we present the relationship between the aggressive treatment style measures and total hospital expenditures based on the surgical population in this study. We found that the Dartmouth aggressiveness measures were associated with overall hospital spending for the procedures we studied. Furthermore, a previous study by Barnato et al. (2009)
has shown a high correlation between a hospital's end-of-life spending and a hospital's spending on all patients who are severely ill, that is, those with a probability of death >21 percent. Finally, and most important, if the aggressiveness measures were meaningful only for the nine conditions used in their measure, we should not have seen any significant relationship between these aggressiveness measures and our outcome measures, yet we found highly significant and large effects for both mortality and failure to rescue.
While we have reported our results in terms of adjusted odds ratios, it is also helpful to think about the absolute probability differences between hospitals with more or less aggressive treatment styles. To do this, we computed directly standardized (Bishop, Fienberg, and Holland 1975
) 30-day death rates and failure-to-rescue rates for all patients in the study, assuming each patient was operated on in a hospital associated with a 25th percentile versus a 75th percentile aggressive treatment style. The mean differences in rates of death were 2.7, 1.6, and 6.7 deaths per 1,000 admissions for general surgery, orthopedics, and vascular surgery, respectively. For failure-to-rescue the rates were 5.6, 4.1, and 9.4 deaths per 1,000 patients with complications, respectively. In populations where the underlying mortality rate is higher, such as in vascular surgery, the significant reduction in the odds of death at hospitals with more aggressive treatment style is reflected in a larger number of reduced deaths than when the underlying risk of death (or failure-to-rescue) is lower.
One can easily imagine that while aggressive care is associated with better survival after surgery, the cost per life saved may be quite variable (Kaestner and Silber, unpublished data, 2009). In a recent study by Chandra and Staiger (2007)
, they find that aggressive hospital care may save lives and do so at a relatively low cost. Other studies report similar results (Card, Dobkin, and Maestas 2008
; Doyle 2005
; Doyle, unpublished data, 2008).
The fact that we found that increased aggressiveness was significantly associated with reduced mortality and reduced failure-to-rescue is an especially important observation given the nature of the definition of aggressiveness. Much has been made of the lack of association between aggressiveness and process measures (Yasaitis et al. 2009
). If aggressiveness was not associated with better process, as the argument goes, then reducing aggressiveness should not worsen quality. Our results suggest that when measuring outcomes in surgical procedures, there appears to be a significant and durable benefit from undergoing surgery in hospitals with a more aggressive treatment style as defined by the Dartmouth Atlas.
Finally, in a recent study of heart failure outcomes in six California Hospitals, Ong et al. (2009)
reported a strong association between increased aggressiveness and better outcomes, in distinction to the Dartmouth results. Ong points out that by defining aggressiveness through looking backward and only studying the deaths, there is an implicit assumption that the probability of death, conditional on severity, is not influenced by the intensity of treatment and furthermore that those who survived had similar expenditure patterns—something that may not be true if hospitals use the patient's chances of survival as a variable in deciding how to expend resources. In a similar manner, we chose to examine the outcomes of surgical patients looking forward in time, and we present findings consistent with Ong, but on a larger scale. Our thought was that the problems inherent in the aggressiveness methodology would be especially transparent in surgical patients because we could directly measure failure-to-rescue, the probability of death after complications. Presumably there are costs to rescuing patients who develop complications, and the choice to be less aggressive may directly influence the failure-to-rescue effectiveness.
Our findings suggest a more narrow application of the concept of “aggressiveness” than the one now widely adopted, as evidenced by articles in popular press and statements by policy makers. We have presented evidence of a beneficial effect of aggressive or intensive treatment of surgical patients, in contrast to the broader view that aggressiveness is wasteful. We believe that these results provide an important cautionary note for a national-level policy based on a one-size-fits-all interpretation of aggressiveness, for example, by reimbursing hospitals by making use of a standard of care that is based on differences in geographic variation in resource use such as the aggressiveness measures reported in the Dartmouth Atlas.
Why does aggressive care appear to be associated with better surgical outcomes? Possibly the more aggressive hospitals are also ones with better facilities for handling patients who develop complications. Note, however, that we did report similar results when we adjusted for the hospital characteristics of size, technology, nurse-to-bed ratio, nurse skill mix, and finally resident-to-bed ratio.
In conclusion, surgery at hospitals with more aggressive treatment styles is associated with better surgical outcomes. Studying these practice styles may help improve quality. The recent emphasis on comparative effectiveness research will hopefully allow us to better understand why patients operated on at more aggressive hospitals appear to have better outcomes.